A network model for capped distance-based tolls
Transcript of A network model for capped distance-based tolls
![Page 1: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/1.jpg)
A network model for capped distance-based tolls
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 2014
Calin D. Morosan
Michael Florian
![Page 2: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/2.jpg)
Capped distance-based tolls
Model formulation
Mathematical properties
Algorithm description
A numerical example
Contents
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 20142
A numerical example
Sydney highway toll study
Conclusions
![Page 3: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/3.jpg)
Distance-based tolls charge the road users proportionally to the distance of their trip, or more general, to the usage of the road
Capped distance-based tolls
Toll cost Toll cost Toll cost Toll cost
on a pathon a pathon a pathon a path
Max tollMax tollMax tollMax toll
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 20143
Path lengthPath lengthPath lengthPath length
Min tollMin tollMin tollMin toll
Max tollMax tollMax tollMax toll
Cost per kmCost per kmCost per kmCost per km
![Page 4: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/4.jpg)
The classical network equilibrium formulation which adds a fixed cost to each link is no longer valid since the tolls are non-additive due to the capping
The conversion into a link-additive network equilibrium model is done via a network construction
Model formulation
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 20144
![Page 5: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/5.jpg)
The classical network equilibrium formulation which adds a fixed cost to each link is no longer valid since the tolls are non-additive due to the capping
The conversion into a link-additive network equilibrium model is done via a network construction
Model formulation
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 20145
a b c d
( )maxmin ,a d a b c dt t t t t t→ = + + +
![Page 6: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/6.jpg)
The classical network equilibrium formulation which adds a fixed cost to each link is no longer valid since the tolls are non-additive due to the capping
The conversion into a link-additive network equilibrium model is done via a network construction
Model formulation
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 20146
a b c d
( )maxmin ,a d a b c dt t t t t t→ = + + +
![Page 7: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/7.jpg)
The network equilibrium problem is formulated on the augmented network A U D, partitioned into three disjoint sets� Links in the base network without tolls A − T
Model formulation
( ),a a ac s v a A T= ∈ −
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 20147
� Links in the base network with tolls T
� Additional links D
�
1
, ,a a a ajk
j a k n
c s v v t j k a T≤ ≤ ≤ ≤
= + + ≠ ∈
∑
� �
�
1
, ,alm jk lmj a k n
m
aa l
c v v t j k lm Ds≤ ≤ ≤ ≤=
+ + ≠ ∈
= ∑∑ ���
![Page 8: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/8.jpg)
The network equilibrium problem is formulated on the augmented network A U D, partitioned into three disjoint sets� Links in the base network without tolls A − T
Model formulation
( ),a a ac s v a A T= ∈ −
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 20148
� Links in the base network with tolls T
� Additional links D
�
1
, ,a a a ajk
j a k n
c s v v t j k a T≤ ≤ ≤ ≤
= + + ≠ ∈
∑
� �
�
1
, ,alm jk lmj a k n
m
aa l
c v v t j k lm Ds≤ ≤ ≤ ≤=
+ + ≠ ∈
= ∑∑ ���
![Page 9: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/9.jpg)
The network equilibrium problem is formulated on the augmented network A U D, partitioned into three disjoint sets� Links in the base network without tolls A − T
Model formulation
( ),a a ac s v a A T= ∈ −
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 20149
� Links in the base network with tolls T
� Additional links D
�
1
, ,a a a ajk
j a k n
c s v v t j k a T≤ ≤ ≤ ≤
= + + ≠ ∈
∑
� �
�
1
, ,alm jk lmj a k n
m
aa l
c v v t j k lm Ds≤ ≤ ≤ ≤=
+ + ≠ ∈
= ∑∑ ���
![Page 10: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/10.jpg)
With the cost vector C(v) defined for each link type, the equivalent optimization problem for the user equilibrium is
Mathematical properties
We showed that this is possible iff the Jacobian matrix of
( )T
0min C d∫
v
x x
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201410
We showed that this is possible iff the Jacobian matrix of the cost vector C(v) is symmetric everywhere
�
�
�
�
��, ,
ij kl
ijkl
s sij D kl D
v v
∂ ∂= ∈ ∈
∂ ∂
�
� �, ,
ija
aij
ssa T ij D
v v
∂∂= ∈ ∈
∂ ∂
![Page 11: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/11.jpg)
The algorithm can be summarized as follows:
� Augment the original network with additional links
� From the toll on base network links compute the toll and define the cost functions on the additional links
� Solve the traffic user equilibrium problem on the augmented network with tolls as fixed link cost
Algorithm description
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201411
network with tolls as fixed link cost
� Accumulate the flows on the additional links and assign it to the toll links
Any method for obtaining UE flows can be used
A parallelized bi-conjugate variant* of the linear approximation method has been implemented in Emme 4.1
*As described in Mitradjieva, M. and Lindberg, P.O. (2013)
![Page 12: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/12.jpg)
A numerical example
Mountain route with no alternative
cost 22
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201412
![Page 13: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/13.jpg)
Mountain route with no alternative
Demand 600 trips from 1 to 2
A numerical example
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201413
![Page 14: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/14.jpg)
A numerical example
Mountain route with no alternative
Demand 1200 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201414
![Page 15: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/15.jpg)
A numerical example
Mountain route with no alternative
Demand 1800 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201415
![Page 16: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/16.jpg)
A numerical example
Mountain route with no alternative
Demand 2400 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201416
![Page 17: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/17.jpg)
A numerical example
Mountain route with toll road alternative
cost 22
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201417
cost: travel time 5 + toll 20
![Page 18: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/18.jpg)
A numerical example
Mountain route with toll road alternative
Demand 2400 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201418
![Page 19: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/19.jpg)
A numerical example
Mountain route with toll road alternative
Demand 3000 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201419
![Page 20: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/20.jpg)
A numerical example
Mountain route with toll road alternative
Demand 3600 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201420
![Page 21: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/21.jpg)
A numerical example
Mountain route with toll road alternative
Demand 4200 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201421
![Page 22: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/22.jpg)
A numerical example
Mountain route with toll road alternative
Demand 4800 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201422
![Page 23: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/23.jpg)
A numerical example
Mountain route with toll road alternative cap $18
Demand 4800 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201423
![Page 24: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/24.jpg)
A numerical example
Mountain route with toll road alternative cap $16
Demand 4800 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201424
![Page 25: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/25.jpg)
A numerical example
Mountain route with toll road alternative cap $15
Demand 4800 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201425
![Page 26: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/26.jpg)
A numerical example
Mountain route with toll road alternative cap $14
Demand 4800 trips from 1 to 2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201426
![Page 27: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/27.jpg)
The method has been applied in Sydney, Australia
� Land area – 4,700 km2
� Population – 4.7 million
� Employment – 2.3 million
� Zones – 2,123
� Nodes – 14,900
Sydney highway toll study
M2
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201427
� Nodes – 14,900
� Links – 41,350
Key features:� different cap values
� ~20km travelled
� congested road
� significant % of trucks
M7
![Page 28: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/28.jpg)
30 classes of traffic corresponding to distributed values of time for the private car and truck modes
Extraction of cordon matrices, economic benefits, etc
4 time periods (AM, IP, PM, NT)
Sydney highway toll study
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201428
VOT
![Page 29: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/29.jpg)
Sydney M7 highway toll study
Uncapped toll� max flow 7000 vph
� average tolled distance 12.0 km
Toll cap AUD 7.20
� max flow 7300 vph
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201429
� max flow 7300 vph
� average tolled distance 12.5 km
![Page 30: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/30.jpg)
Sydney M7 highway toll study
Uncapped toll� max flow 7000 vph
� average tolled distance 12.0 km
Toll cap AUD 3.60
� max flow 8200 vph
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201430
� max flow 8200 vph
� average tolled distance 14.4 km
![Page 31: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/31.jpg)
different cap values were used, varying from 0$ to 15$
Sydney M2 highway toll study
link A
link B
link C
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201431
link D
![Page 32: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/32.jpg)
This method has the advantage that it can be solved by any algorithm that computes user equilibrium flows
Previous contributions addressing UE flows with non-additive tolls resort to path enumeration or are restricted to uncapped linear dependent distance toll
This method is easy to apply and does not require post
Conclusions
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 201432
This method is easy to apply and does not require post processing that uses approximations
This method has also been extended to treat ramp-to-ramp tolls
Thanks to Jacobs/SKM for providing sample data and feedback
![Page 33: A network model for capped distance-based tolls](https://reader031.fdocuments.in/reader031/viewer/2022030220/621e8f13806c9f62253acdfc/html5/thumbnails/33.jpg)
The Evolution of Transport Planning
Innovations in Travel Modeling, Baltimore, MD, April 27-30, 2014332007